A statistical reliability analysis of the variability and upper percentiles of the wood strand thickness of oriented strand board

Oriented strand board (OSB) is an important structural engineered wood product used predominately in housing construction, with OSB revenue of around $14 billion in 2005. OSB is a product with a low environmental impact or "carbon footprint." In this thesis, reliability and statistical tools are applied to gain insights of the strand thickness for OSB panels manufactured in the Eastern United States. The thesis, also, develops new techniques to more realistically estimate upper percentiles via induced left censoring. An OSB panel consists of thousands of resinated wood strands that are formed in mats of oriented strands and pressed with heat causing thermal-activated bonding. The variability of OSB strand thickness for six manufacturers is examined. Strand thickness variability has been documented in the literature as having a strong influence on mat formation quality and subsequently the strength properties of OSB wood panels.However, there is an absence in the literature of assessing strand thickness variability from OSB mills. The goals of the thesis are to quantify and characterize strand thickness, plus apply reliability techniques, such as Kaplan-Meier curves and left censoring, to better characterize the probability and percentiles of strand thickness. The thesis further explores graphically and statistically the thickness of the strands through histograms, probability plots, box plots, and so on. Using induced percentile left censoring for improved model fitting, bootstrapping methods are employed for better estimating the upper percentiles, which are of particular interest due to their importance in the manufacturing process. If the OSB strands are too thick, machines and presses can be damaged at great expense. A comparison of the upper percentiles for six OSB mills identifies mills at greater risk for equipment damage and financial loss.Left percentile censoring is explored and used in conjunction with bootstrapping to calculate confidence intervals for the upper percentiles. Appropriate parametric models are used for the bootstrapping and nonparametric bootstrapping methods are presented as a means of comparison. Better estimation of upper percentiles promotes continuous improvement of preventive maintenance and product quality. Continuous improvement has never been more important for manufacturers than it is now given the severely constrained housing markets and the economic recession of 2009.